Crossing distances of surface-knots

Topology and its Applications 154 (2007) 1532–1539
www.elsevier.com/locate/topol
Crossing distances of surface-knots
Tsukasa Yashiro a,b
a Department of Mathematics, University of Auckland, 38 Princes St., Auckland, New Zealand
b Department of Mathematics and Statistics, Sultan Qaboos University, PO Box 36, PC 123 Al-Khod, Sultanate of Oman
Received 25 May 2005; received in revised form 10 April 2006; accepted 10 April 2006
Abstract
A surface-knot is an embedded closed connected oriented surface in 4-space. A surface diagram is a projection of a surface-knot
into 3-space with crossing information. In this paper we define a distance from a special surface diagram to a trivial diagram as the
minimal number of special double cycles, where we can change the crossing information to obtain the trivial diagram. We estimate
the distance using the number of 1-handles needed to obtain a trivial diagram.
© 2006 Elsevier B.V. All rights reserved.
MSC: primary 57Q45; secondary 57M25
Keywords: Surface-knot; Distance; Trivial surface
1. Introduction
A surface-knot is a smoothly embedded closed connected oriented surface in R4 . A surface-knot F is equivalent
to a surface-knot F if F is isotopic to F in R4 . We denote this by F ∼ F . A trivial surface is a surface in R4 ,
which is equivalent to the boundary of a handlebody embedded in R3 × {0} [5]. We denote a trivial surface by F. Let
π : R4 → R3 be the orthogonal projection defined by π(x1 , x2 , x3 , x4 ) = (x1 , x2 , x3 ). Adding crossing information on
the crossing set of π(F ), we obtain a diagram called a surface diagram [2], which will be denoted by DF . If two
diagrams DF and DF are of equivalent surface-knots F and F , then we denote this by DF ∼ DF . A diagram DF
of a trivial surface F is called a trivial diagram. The projected image π(F ) of a surface-knot F without the crossing
information will be denoted by |DF |. If DF ∼ DF and |DF | = |DF |, then we denote this by DF = DF . It is known
that every immersed circle in R2 can be lifted into R3 as a knot. It is proved that there are some immersed surfaces in
R3 , which cannot be lifted into R4 [4,3,11]. It is not known whether or not every liftable generic surface is realised as
a trivial surface in R4 . Some surface-knots have diagrams, each of which has special double curves; we can change
the crossing information of them to obtain trivial diagrams. Such a deformation forms a track between given the
surface diagram DF and its trivial diagram DF such that |DF | = |DF |. The minimal number of the crossing change
operations will indicate the difference between F and F. Let B n be a closed unit n-dimensional ball in Rn , and let
B m (m < n) be the set of points in B n whose last (n − m) coordinates are zero. Let N (F ) ∼
= F × B 2 be a closed
4
4
tubular neighbourhood in R with the projection p : N (F ) → F . Let X = cl(R \ N (F )), where cl means the closure.
E-mail addresses: yashiro@math.auckland.ac.nz, yashiro@squ.edu.om (T. Yashiro).
0166-8641/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.topol.2006.04.028
T. Yashiro / Topology and its Applications 154 (2007) 1532–1539
1533
Fig. 1.
There is a map q : N(F ) → B 2 corresponding to the projection onto the B 2 factor. We fix such a map q : N (F ) → B 2 .
Consider an embedding h : [0, 1] × B 3 → X such that
(1) h([0, 1] × {0}) is a properly embedded arc with an orientation induced from the natural orientation of [0, 1],
(2) h([0, 1] × B 3 ) is a closed tubular neighbourhood of h([0, 1] × {0}) in X and
(3) q(h({i} × B 2 )) is a single point, p(h({i} × B 2 )), (i = 0, 1) are disjoint discs and induce opposite orientations
on F .
There are canonical annuli Ai (i = 0, 1) in N (F ) such that ∂Ai = h({i} × ∂B 2 ) ∪ p(h({i} × ∂B 2 )), p(Ai ) = p(h({i} ×
∂B 2 )) and q(Ai ) is a ray in B 2 (i = 0, 1). The surface-knot F + h is defined as the following:
F + h = F \ p h {0, 1} × B 2 ∪ A0 ∪ A1 ∪ h [0, 1] × ∂B 2 ,
(1)
(see [1,9]). We call the cylinder A0 ∪A1 ∪h([0, 1]×∂B 2 ) a 1-handle denoted by h. We will also use h for the projected
image π(h) and for this case, we call h a 1-handle attached to DF . In the following, we will distinguish two kinds of
1-handles in R3 depicted in Fig. 1. Note that these 1-handles are deformed into each other by isotopy deformations.
Attaching a 1-handle is described as a sequence of deformations of surface with a critical point called a hyperboloidal
transformation [9]. A 1-handle h is trivial if h is obtained by the hyperboloidal transformation along an arc δ such
that π(δ) is trivial in R3 \ π(F ). It is proved that there are some 1-handles h1 , . . . , hn such that F + (h1 ∪ · · · ∪ hn ) is
trivial [5]. The minimum number of such 1-handles is a surface-knot invariant and is called an unknotting number of F
denoted by u(F ) [5]. In this paper we describe relations between the number of crossing changes and the unknotting
number (Theorem 2.1).
The organisation of this paper is the following. Section 2 will describe cross change operations for surface-knots
and give our main result. In Section 3 we will introduce Roseman moves, which enable us to visualise isotopy deformations of surface-knots. Section 4 describes how cross change operations are realised by attaching handles. Finally,
we will prove Theorem 2.1.
2. Cross change operations
We can assume that the projected image of a surface-knot F is a generic surface in R3 . The image π(F ) has either
a regular point or a double point or a triple point or a branch point [9]. Let h : R4 → R be the projection defined
by h(x1 , x2 , x3 , x4 ) = x4 . For a double point p in DF , there is a 3-ball neighbourhood B(p) of p in R3 such that
π −1 (B(p)) ∩ F consists of two disjoint discs D− and D+ , and for x ∈ D− and y ∈ D+ , h(x) < h(y). Then we
call D− and D+ the lower sheet and the upper sheet at p respectively. For a triple point q in DF , there is a 3-ball
neighbourhood B(q) of q in R3 such that π −1 (B(q)) ∩ F consists of three disjoint discs DB , DM and DT and for
x ∈ DB , y ∈ DM and z ∈ DT , h(x) < h(y) < h(z). Then DB , DM and DT are called the bottom sheet, the middle
sheet and the top sheet respectively. There are three types of double edges in DF connecting to q; π(DB ) ∩ π(DM ),
π(DM ) ∩ π(DT ) and π(DB ) ∩ π(DT ), called type (b/m), (m/t) and (b/t) respectively [11].
Let c be a closed double curve in a surface diagram DF of a surface-knot F . Viewing c as an immersed circle
in |DF |; that is, the image of an immersion α : [0, 1] → c with α(0) = α(1), then the pre-image of triple points on
α([0, 1]) is an ordered sequence 0 < t1 < · · · < tu < 1 denoted by T (c). For a point ti ∈ T (c), we define the type of ti
is the type of the double edge on α([0, 1]) at α(ti ). A double curve c is called a cross-exchangeable cycle if there exists
an ordered sequence T (c) = {t1 , . . . , tu } of the pre-image of triple points along c such that it satisfies the following
conditions:
(1) If there exists ti ∈ T (c) (1 i u) such that α −1 (α(ti )) = ti , then the type of ti is (b/m) or (m/t), and
1534
T. Yashiro / Topology and its Applications 154 (2007) 1532–1539
Fig. 2.
(2) (2a) if there exist ti , tj ∈ T (c) such that α(ti ) = α(tj ) (1 i < j u), then types of {ti , tj } are either
{(b/m), (b/t)} or {(m/t), (b/t)} in the order; or
(2b) if there exist ti , tj , tk ∈ T (c) such that α(ti ) = α(tj ) = α(tk ) (1 i < j < k u), then types of {ti , tj , tk }
are either {(b/m), (b/t), (m/t)} or {(m/t), (b/t), (b/m)} in the order.
Let F be a surface-knot and let DF be a surface diagram of F . Assume that DF has a cross-exchangeable cycle
c ⊂ DF and p ∈ c then there are the lower sheet D− and the upper sheet D+ at p. We introduce an operation at p
defined by Giller [4]. Push D+ down into D− so that a pair of crossing points p+ , p− ∈ R4 are created. We denote the
projected points π(p+ ) and π(p− ) by p+ and p− respectively. This operation gives an immersed surface in R4 . We
will call this operation a push down operation at p ∈ c. In the modified diagram p− can move along c and p− passes
triple points so that p− reaches p+ from the other side. Applying the reverse operation of the push down operation, we
can eliminate those crossing points. This operation creates a cross-exchangeable cycle c . We call the operation a cross
change operation along c. We denote the resulting diagram by DF (c). If there are some disjoint cross-exchangeable
cycles {c1 , . . . , cn } in DF , then. We denote the DF (c1 , . . . , ck−1 )(ck ) by DF (c1 , . . . , ck ) (2 k n).
Let Fg be the set of surface-knots of a fixed genus g 0, each of which has a diagram DF with disjoint crossexchangeable cycles {c1 , . . . , cn } such that DF (c1 , . . . , cn ) = DFg , where Fg is a trivial surface of genus g. Define a
distance d2 (F, Fg ) by the minimal number of such cross-exchangeable cycles. Then we have the following.
Theorem 2.1. Let F be a surface-knot in Fg and let u(F ) be the unknotting number of F . Then the following holds:
u(F ) d2 (F, Fg ) + ,
(2)
where is the number of non-simple cross-exchangeable cycles in the minimal diagram giving d2 (F, Fg ). In particular,
u(F ) 2d2 (F, Fg ).
Note that S. Kamada obtained a similar result with surface braids [7,8]. The following corollary is an immediate
consequence of Theorem 2.1.
Corollary 2.1. For every pair (n, g) of non-negative integers, there exists a surface-knot F ∈ Fg such that
n d2 (F, Fg ) 2n.
(3)
Proof. From Theorem 2.1, u(F ) 2d2 (F, Fg ) for any F ∈ Fg . Let S be a ribbon 2-knot obtained by a fusion as in
Fig. 2. It is proved u(S) = 1 [5]. Let Sn = ni=1 Si , (Si ∼ S), where is meant by the connected sum. Then F. Hosokawa,
T. Maeda and S. Suzuki proved that u(Sn ) = n [6]. On the other hand, Sn has a diagram DSn with disjoint simple
cross-exchangeable cycles {c1 , . . . , cn } such that DSn (c1 , . . . , cn ) = DS2 , where S2 is a trivial sphere. Thus for a nonnegative integer n, n d2 (S2n , S2 ) 2n. We can attach g copies of trivial 1-handles to S2n to obtain a surface of
genus g. 2
3. Roseman moves
D. Roseman [10] introduced seven types of local deformations and he proved the following lemma.
Lemma 3.1 (D. Roseman). Let F1 and F2 be surface-knots. F1 is equivalent to F2 if and only if DF1 is deformed into
DF2 by a finite sequence of seven types of local deformations depicted in Figs. 3 and 4.
T. Yashiro / Topology and its Applications 154 (2007) 1532–1539
1535
Fig. 3.
Fig. 4.
In Figs. 3 and 4, the deformation from the left diagram to the right diagram is called an Ri+ move and the other
direction is called an Ri− move (i = 1, . . . , 6). For R7 moves we do not distinguish the directions. For the terminal
diagram of the R4+ move, the diagram, with a double segment bounded by branch points, we will call it a bug. Note
that in R6± moves at the triple point, the double edge bounded by the branch point and the triple point is either of type
(b/m) or of type (m/t).
4. Attaching handles
Let F be a surface-knot and let DF be a surface diagram of F . Let c be a cross-exchangeable cycle in DF . In
this section we will describe observations on deformations of double curves and attaching handles. First, we attach
a 1-handle at a double line on a model of some double lines and triple points and then deform it. Using this we will
prove Lemma 4.1 and Corollary 4.1. Then we look at attaching two 1-handles at the double line on the model and
deform it to create another double line having the opposite crossing information. Using this observation, we will prove
Lemma 4.2.
Let P1 be the plane {(x, 0, z, 1) ∈ R4 } and let P2 be the plane {(x, y, 0, 0) ∈ R4 }. We denote the half spaces
{(x, y, z, t) ∈ R4 ; x > 0} and {(x, y, z, t) ∈ R4 ; x < 0} by R4+ and R4− respectively. We denote the double curve
π(P1 ) ∩ π(P2 ) by c. Now we add two planes P3 = {(−1, y, z, 2) ∈ R4 } and the plane P4 = {(−2, y, z, −1) ∈ R4 } to
the diagram D to obtain two triple points t1 , t2 on c. The diagram π(P1 ∪ P2 ∪ P3 ∪ P4 ) with crossing information
will be denoted by D.
Take two points p1 = (0, 0, 1, 1) ∈ P1 and p2 = (0, 1, 0, 0) ∈ P2 . Join p1 and p2 by a simple arc δ so that π(δ) ∩
(π(P1 ) ∪ π(P2 )) = {π(p1 ), π(p2 )}. Apply the hyperboloidal transformation along δ to obtain an attached 1-handle h
to D (see Fig. 5). Let Q be the plane {(0, y, z) ∈ R3 }. Then (D + h1 ) ∩ Q is depicted in the left of Fig. 6. Thus we
Fig. 5.
Fig. 6.
1536
T. Yashiro / Topology and its Applications 154 (2007) 1532–1539
Fig. 7.
can see the shaded disc, in which we can apply an R5+ move so that we obtain a pair of branch points b1 ∈ π(R4+ ) and
b2 ∈ π(R4− ).
Types of triple points t1 and t2 along c are (b/m) and (m/t) respectively. This implies that the branch points created
by the above operation can pass P3 and P4 by R6− moves and thus t1 and t2 are eliminated. From this observation, we
have the following lemma and its corollary.
Lemma 4.1. Let F be a surface-knot and let DF be a surface diagram of F . Let c ⊂ DF be a simple crossexchangeable cycle. Let p ∈ c be a double point. Assume that there exists an embedded disc E ⊂ R3 such that
(1) E ∩ |DF | = γ is a simple loop based at p,
(2) γ does not meet either double points or triple points or branch points, and
(3) E is transversal to DF along ∂E.
Then DF (c) = DF .
Proof. Since the disc E exists, we can apply an R5+ move along E so that we obtain a pair of branch points {b1 , b2 }
on c; note that after this operation, the double circle becomes an arc bounded by b1 and b2 . Since c is simple crossexchangeable cycle, we can move b2 along c and pass through triple points on c to approach b1 from the other side.
Then we obtain a bug with a handle attached on its back (Fig. 7). One can change the crossing information of the
bug by rotating the bug part to switch b1 and b2 . Then move b1 instead of b2 to recover the double curve with the
switched crossing information. Apply an R5− move to obtain DF (c). All local deformations used in the above are
isotopy deformations. By Lemma 3.1, DF (c) ∼ DF and thus DF (c) = DF . 2
We obtain the following.
Corollary 4.1. Let F be a surface-knot and let DF be a surface diagram of F . Assume that there is a simple crossexchangeable cycle c ⊂ DF . Then DF + h1 = DF (c) + h1 .
Proof. For the cross-exchangeable cycle c, we attach the 1-handle at c, we obtain a loop γ based at a point p ∈ c so
that γ bounds an embedded disc E with E ∩ |DF | = ∂E. This satisfies conditions of Lemma 4.1. Thus DF (c) + h1 =
DF + h1 . 2
The second model of attaching 1-handles is the following. The symbols P1 , P2 , P3 , P4 , D and Q mean as in the
first model. Take two points q1 = (0, 0, 1, 1), q2 = (0, 0, −1, 1) ∈ P1 . Join q1 and q2 by a simple arc δ = δ1 so
that π(δ) is in Q. δ ∩ π(P2 ) is a single point such that at the crossing point, the height of δ is lower than P2 .
Along δ apply the hyperboloidal transformation to obtain an attached 1-handle. We denote the resulting diagram
by D + h11 (see the right picture of Fig. 8). The diagram of (D + h11 ) ∩ Q is depicted as the left picture of Fig. 8.
We attach a 1-handle h12 to D + h11 as in the middle picture of Fig. 9. Then we can see that there are two discs, in
which we can apply R5+ moves to obtain pairs of branch points {b1 , b2 } and {b3 , b4 } such that b1 , b3 ∈ π(R4+ ) and
b2 , b4 ∈ π(R4− ); there are no branch points on Q (see the right picture of Fig. 9). We can assume that all branch points
are in π(D) ∩ π({(x, y, z, t) ∈ R4 | |x| 2.5, |y| 1, |z| 1.5}); the restricted diagram will be denoted by D .
T. Yashiro / Topology and its Applications 154 (2007) 1532–1539
1537
Fig. 8.
Fig. 9.
Fig. 10.
The type of t1 along c is (b/m) while the type of t2 is (m/t). Consider the partial diagram cl(D ∩ π(R4− )). Take
two double curves γ1 and γ2 such that γ1 is on c starts from the branch point b2 and γ2 starts from a point in the
boundary Q ∩ cl(D ∩ π(R4− )) bounded by b4 (see the very left picture of Fig. 10). Since the type of t1 is (b/m), then
the branch point b2 can pass the triple point and eliminate the triple point (see Fig. 10). The very left picture of Fig. 10
is a partially cut diagram between planes Q and π(P3 ). The second left picture of Fig. 10 shows the projection image
of the singularity set of π(D) to {(x, y) ∈ R2 | −2.5 x 0, |y| 1}. We also move the arc γ2 so that γ2 and P3
create two extra triple points. Along γ2 the types of those triple points are both (b/m), thus the branch point b4 can
pass P3 so that one of the triple points is eliminated. The resulting triple point will be denoted by t1 . Similarly, these
branch points can pass P4 . Thus the operation has created t1 and the new triple point t2 on the deformed γ2 (see the
very right picture of Fig. 10).
Note that the crossing information along the deformed γ2 differs from that of γ1 . Also the above operation will
replace triple points {t1 , t2 } with {t1 , t2 }. From this observation, we obtain the following.
1538
T. Yashiro / Topology and its Applications 154 (2007) 1532–1539
Fig. 11.
Fig. 12.
Lemma 4.2. Let F be a surface-knot and let DF be a surface diagram of F . Assume that DF contains a crossexchangeable cycle c with some self-intersections. Then there exist 1-handles h11 , h12 , h¯ 11 and h¯ 12 such that DF + (h11 ∪
h12 ) ∼ DF (c) + (h¯ 11 ∪ h¯ 12 ).
Proof. Attach 1-handles h11 and h12 to DF at c as we saw in the second observation. It is not difficult to see that there
is a partial diagram, which is homeomorphic to D with branch points b1 , b2 , b3 and b4 . These branch points are
obtained as in the second observation. Apply the moving operation of branch points. Since c is a cross-exchangeable
cycle, there are an immersion α : [0, 1] → c with α(0) = α(1) and the ordered sequence T (c) = {t1 , . . . , tu }. There
are innermost multiple points of α in T (c). We may consider the deformation for this case without loss of generality.
There are two cases.
(1) There are ti , tj ∈ T (c) such that t = α(ti ) = α(tj ) (1 i < j u). Types of {ti , tj } are either {(b/m), (b/t)}
or {(m/t), (b/t)} in the order. Assume that the type of ti is (b/m). As we saw in the second observation, if we apply
the operation along c, t is created near t and then t is eliminated. The deformation approaches t as the deformation
continues, then the type of t with respect to the order is (m/t). Then t will be created near t (see Fig. 11). The
type of t is (m/t) thus t can be eliminated by R6− move. Similarly, if the types of {ti , tj } are {(m/t), (b/t)}, then the
operation can continue over those triple points.
(2) There are ti , tj , tk ∈ T (c) such that t = α(ti ) = α(tj ) = α(tk ) (1 i < j < k u). Types of {ti , tj , tk } ⊂ T (c)
are either {(b/m), (b/t), (m/t)} or {(m/t), (b/t), (b/m)} in the order. From the above argument, for either case, the
operation can be done at the triple point t creating t and eliminating t and then the operation is applied to t creating
t and eliminating t (see Fig. 11). Finally the operation is applied to t . This can be done, since at t , the type is
(b/m) or (m/t). Thus the operation creates a triple point t near t , then t is eliminated.
From the above arguments, branch points b2 and b4 with double arcs can travel along c and approach b1 and
b3 from the other side. Now we complete the operation by eliminating four branch points as follows. When b2 and
b4 approach b1 and b3 , the double arcs are depicted in the left picture of Fig. 12. Let Q be a vertical plane in R3
separating {b2 , b4 } and {b1 , b3 } in the partial diagram D . The left picture in Fig. 13 shows the intersection of D and
a thin neighbourhood of Q in R3 . We can see new 1-handles h¯ 11 and h¯ 12 (shaded part in pictures). Applying an R7
move to join the double curves connecting to b3 and b4 . Then we obtain a cross-exchangeable cycle c with triple
points created by the above operations. Slide the handle h¯ 12 down under the horizontal sheet (see the middle picture
of Fig. 13). This operation creates two double segments, which are bounded by {b1 , b3 } and {b2 , b4 }. Note that at
this stage two bugs with double segments are isolated. Then an R4− move can be applied to eliminate these double
segments. The handle h¯ 12 seems to link with the horizontal sheet (see the middle picture of Fig. 13), but after bugs
are eliminated, it is obvious that there is a disc satisfying the condition of Lemma 4.1. Thus we can unlink the handle
T. Yashiro / Topology and its Applications 154 (2007) 1532–1539
1539
Fig. 13.
from the horizontal sheet (see the right picture of Fig. 13). Now the diagram DF (c) with 1-handles h¯ 11 and h¯ 12 appears.
Thus the resulting diagram is DF (c) + (h¯ 11 ∪ h¯ 12 ). This completes the proof. 2
5. Proof of Theorem 2.1
Proof. Let n = d2 (F, Fg ). There is a surface diagram DF of F such that DF has cross-exchangeable cycles
{c1 , . . . , cn } ⊂ DF such that DF (c1 , . . . , cn ) = DFg .
Let be the number of non-simple cycles of {c1 , . . . , cn } and put m = n − . Changing indices of {c1 , . . . , cn }, we
may assume that ci (i m) are simple and cj (m + 1 j ) are non-simple. For each ci (i m), there is a 1-handle
h1i with DF + h1i = DF (c1 , . . . , cn ) + h1i by Lemma 4.1. Thus
DF +
m
h1i = DF (c1 , . . . , cm ) +
i=1
m
h1i .
(4)
i=1
For each cj (m + 1 j ), by Lemma 4.2 there are handles h1j 1 , h1j 2 , h¯ 11 and h¯ 12 such that DF + (h1j 1 ∪ h1j 2 ) ∼ DF (cj ) +
(h¯ 1j 1 ∪ h¯ 1j 2 ). Thus
DF +
m+
m+
1
hj 1 ∪ h1j 2 ∼ DF (cm+1 , . . . , cm+ ) +
h¯ 1j 1 ∪ h¯ 1j 2 .
j =m+1
(5)
j =m+1
From the assumption DF (c1 , . . . , cn ) = DFg . The attached handles h¯ 1j 1 and h¯ 1j 2 (m + 1 j ) can be deformed into
trivial handles [5]. Therefore, we need to attach (n + ) 1-handles to F to obtain a trivial surface Fg+n+ . This implies
that u(F ) n + = d2 (F, Fg ) + . 2
Acknowledgements
The author is supported by JSPS post doctoral fellowship for foreign researchers. He would like to thank to Professor S. Kamada and Professor A. Kawauchi for giving him helpful suggestions for the early version of this paper.
References
[1] F. Boyle, Classifying 1-handles attached to knotted surfaces, Trans. Amer. Math. Soc. 306 (1988) 475–487.
[2] J.S. Carter, M. Saito, Knotted Surfaces and Their Diagrams, Mathematical Surveys and Monographs, vol. 55, American Mathematical Society,
Providence, RI, 1998.
[3] J.S. Carter, M. Saito, Surfaces in 3-space that do not lift to embeddings in 4-space, in: Knot Theory, in: Banach Center Publications, vol. 42,
Polish Acad. Sci., Warzawa, 1998, pp. 29–47.
[4] C. Giller, Towards a classical knot theory for surfaces in R4 , Illinois J. Math. 26 (1982) 591–631.
[5] F. Hosokawa, A. Kawauchi, Proposals for unknotted surfaces in four-spaces, Osaka J. Math. 16 (1979) 233–248.
[6] F. Hosokawa, T. Maeda, S. Suzuki, Numerical invariants of surfaces in 4-space, Math. Sem. Notes Kobe Univ. 7 (2) (1979) 409–420.
[7] S. Kamada, Crossing changes for singular 2-dimensional braids without branch points, Kobe J. Math. 13 (2) (1996) 177–182.
[8] S. Kamada, Unknotting immersed surface-links and singular 2-dimensional braids by 1-handle surgeries, Osaka J. Math. 36 (1999) 33–49.
[9] A. Kawauchi, T. Shibuya, S. Suzuki, Descriptions on surfaces in four-space II, Math. Sem. Notes Kobe Univ. 11 (1983) 31–69.
[10] D. Roseman, Reidemeister-type moves for surfaces in four-dimensional space, in: Knot Theory, in: Banach Center Publications, vol. 42, 1998,
pp. 347–380.
[11] S. Satoh, Lifting a generic surface in 3-space to an embedded surface in 4-space, Topology Appl. 106 (1) (2000) 103–113.